Semiconductor Industry Cycle: Historical Patterns and Future Trends

01The Core Cycle Patterns and Driving Mechanisms of the Semiconductor IndustryThe cyclical fluctuations in the semiconductor industry are driven by a dynamic coupling effect of multiple factors:Weakening and Maturation of Technology Drivers: The evolution speed of Moore’s Law has slowed from doubling every year between 1965-1969 to doubling every three years from 2009-2018. Meanwhile, the industry’s growth rate has decreased from 15% between 1990-2000 to 1.7% between 2000-2010, gradually converging towards the global GDP growth rate (approximately 3%). This marks the transition of the industry from a high-volatility, technology-driven phase to a relatively mature, demand-driven phase.Dynamic Balance Mechanism of Supply and Demand: The essence of the semiconductor cycle is the process of supply and demand rebalancing: demand growth → price increase → capital expenditure increase → capacity expansion → supply surplus → price decrease → capacity contraction → supply reduction → demand recovery. This cycle has been fully played out nine times between 1980-2022.Geopolitical and Black Swan Events Impact: External shocks such as the trade friction in 2018 and the COVID-19 pandemic in 2020 have intensified cycle fluctuations. For example, trade friction led to a significant decline in the median revenue growth rate of leading U.S. semiconductor companies, while the contradiction between reduced automotive chip orders at the beginning of the pandemic and the surge in demand for the “stay-at-home economy” later gave rise to the “chip shortage” in 2021.There are three interwoven cycles in the semiconductor industry that jointly determine the trajectory of industry fluctuations, and the driving factors of these cycles evolve over time:1. Three Core Cycles

Innovation Cycle: “Killer Applications” drive growth, such as the proliferation of smartphones from 2000-2010, which boosted semiconductor sales from $139 billion to $271.2 billion (an average annual growth of 8%).Capacity Cycle: The construction of wafer fabs takes 1.5-2 years, and investment decisions are often lagging; a capital expenditure growth rate exceeding 27% should be monitored for potential surplus, while exceeding 56% may trigger severe overcapacity.Inventory Cycle: Following the cycle of “active restocking → passive restocking → active destocking → passive destocking” amplifies price fluctuations.2. Evolution of Cycle Driving FactorsWeakening of Technology Drivers: Moore’s Law has slowed from “doubling every year” (1965-1969) to “doubling every three years” (2009-2018), with the industry’s growth rate decreasing from 15% between 1990-2000 to 1.7% between 2000-2010, converging towards the global GDP growth rate (approximately 3%), entering a “demand-driven mature phase”.Dynamic Balance of Supply and Demand: Between 1980-2022, the cycle of “demand growth → price increase → capacity expansion → surplus → price decrease → contraction” has been completed nine times.External Shocks Intensifying Fluctuations: The trade friction in 2018 led to a decline in U.S. company revenues, while the pandemic initially suppressed demand before surging (initial order cuts → later “stay-at-home economy” → chip shortage in 2021).02Historical Review of the Semiconductor Industry Cycle (1980-2024)Divided into four stages based on core driving forces, experiencing multiple rounds of cyclical fluctuations and shifts in dominance:1. 1980-2000: The PC Era and U.S.-Japan CompetitionJapan’s Rise (1980-1985): With a vertically integrated model of “electronic groups + affiliated banks” and the government VLSI plan, Japan’s DRAM market share surpassed that of the U.S., triggering a price war (a 60% drop in one year); the 1986 U.S.-Japan Semiconductor Agreement temporarily resolved conflicts, but U.S. IDM manufacturers were harmed.U.S. Transformation (1986-2000): Intel exited DRAM to focus on CPUs, while TI concentrated on DSP/analog chips; TSMC was established in 1987, promoting a “Fabless + Foundry” vertical division of labor ecosystem.2. 2000-2010: Internet Bubble and Mobile RevolutionBubble Burst (2000-2002): The internet bubble caused a global semiconductor sales crash, with the Philadelphia Semiconductor Index (SOX) dropping from 1362.1 points to a low (over a 45% decline); the DRAM industry consolidated into three oligopolies: Samsung, Micron, and SK Hynix.Mobile-Driven Recovery (2003-2010): The 2007 iPhone ushered in the mobile internet era, driving demand for communication chips (Qualcomm) and NAND storage; DRAM sales increased by 76.8% in 2007, forming two small cycles from 2004-2010.3. 2010-2022: Multipolar Applications and “Super Cycle”Diverse Demand Supporting Growth: Smartphones (over 1.4 billion units shipped annually), data centers (cloud computing), and automotive electronics (accounting for 40%) surged; sales reached $573.2 billion in 2021 (a record), with the automotive chip shortage driving a 33.1% increase in analog chips.Pandemic Disruption of Cycles (2020-2022): Early 2020 saw capacity contraction, while 2021 experienced a surge in demand for the “stay-at-home economy” leading to a chip shortage; inflation and weak demand in 2022 caused a decline, but annual sales still reached a record high of $574 billion.4. 2022-2024: AI-Driven Recovery (End of the Ninth Cycle)Downturn Period (H2 2022-2023): Sales in 2023 are expected to decline by 10.3% (to $515 billion), with the industry currently facing “four highs”: high inventory (Micron’s turnover days at 214 days), low capacity utilization (60-70% for 8-inch fabs), low equipment spending (down 16.8%), and low silicon wafer demand (down 0.6%).AI-Driven Recovery (2024): Generative AI is driving demand for computing power, with sales expected to reach $627.6 billion (a year-on-year increase of 19.1%); NVIDIA’s data center revenue skyrocketed, and AMD’s related business grew by 94%, with HBM (high bandwidth memory) in short supply.03Characteristics of Three Complete Cycles from 2015-2024

04Structural Changes in the Semiconductor Industry1. Trend of Industry MaturationSlowing Growth + Smoother Fluctuations: The growth rate has decreased from 26% in the 1980s to below 8% after 2000, with the correlation coefficient with global GDP rising from 0.3 to 0.8; after 2001, the amplitude of cycles has narrowed, without severe shocks like the 17% recession in 1985.Rationalization of Capital Expenditure: After 2000, investment deviations have narrowed, with capacity utilization stabilizing at 80-90% (fluctuating between 60-100% before 2000).2. Evolution of Business Model Differentiation

TSMC’s mass production of 3nm in 2023 drives a leap in AI chip performance, becoming the core support of the vertical division of labor model.05Outlook for the Semiconductor Industry from 2025-2035: Innovation, Reconstruction, and Strategic ResilienceThe global semiconductor industry is entering a new decade driven by multiple forces and deep structural reconstruction. Based on historical cycle patterns and current technological inflection points, the evolution of the industry from 2025-2035 will not only be limited to scale expansion but will also present a systematic transformation of growth logic, technological paradigms, and global patterns. The following analysis is deepened from four dimensions: core driving forces, technological paradigm shifts, supply chain resilience, and corporate strategy:1. Core Driving Forces: From “Single Point Breakthrough” to “Multi-Ecological Collaboration”Future industry growth will be supported by three engines: AI, smart vehicles, and edge computing, forming a layered demand structure:Large model parameters reaching trillions will drive the iteration of dedicated AI chips (such as GPUs, TPUs) and stimulate demand for HBM (high bandwidth memory). It is expected that by 2030, the AI chip market will reach $120 billion, with a compound annual growth rate (CAGR) of over 50% for the HBM market from 2023-2030.AI smartphones, AI PCs, and other edge devices will see accelerated penetration, with global AI smartphone penetration expected to reach 43% by 2027, creating demand for low-power, highly integrated SoC chips.Level 4 autonomous vehicle chips will exceed 3,000 units per vehicle, with a value of $2,000 per vehicle. Chip demand will shift from “general processors” to “heterogeneous computing clusters” (such as domain controllers and sensor fusion modules), driving a surge in demand for automotive-grade MCUs and silicon carbide (SiC) power devices.Global IoT connections are expected to exceed 50 billion by 2030, with industrial automation driving demand for edge AI chips and high-reliability sensors; service robots will require computing power 1,000 times higher than current levels, driving innovation in dedicated edge computing chips.Key Turning Point: From 2025-2027, AI computing power demand will spread from the cloud to the edge, and smart driving chips will enter the mass production window; after 2030, the industrial IoT may become a new growth pillar.2. Technological Paradigm Shift: From “Process Miniaturization” to “System-Level Innovation”As the physical limits of Moore’s Law approach, the focus of innovation shifts from increasing transistor density to collaborative breakthroughs in architecture, packaging, and materials:Technologies such as CoWoS and 3D packaging enhance overall computing power through heterogeneous integration. TSMC plans to expand the area of CoWoS substrates to 100×100mm, accommodating 12 HBM4 stacks.The Chiplet market is growing at an annual rate of 40%, reducing chip development costs through modular design and promoting the adoption of small chip interconnect standards (such as UCIe).Silicon carbide (SiC) is accelerating its transition to 8-inch wafers, with STMicroelectronics and ON Semiconductor planning to mass-produce 8-inch SiC products by 2025, enhancing electric vehicle efficiency.Silicon photonic chip technology is maturing, with TSMC planning to launch a 1.6T optical engine in 2025 to address high-speed interconnect bottlenecks within data centers.Quantum computing chips are entering the scale testing phase, with IBM planning to release a 1386-qubit processor in 2025.The RISC-V architecture is penetrating high-performance fields, with the number of cores expected to reach 80 billion by 2025, promoting the open-source hardware ecosystem.Paradigm Impact: Technological competition is shifting from a single process node to a full-stack capability of “process + packaging + architecture”, requiring companies to layout multidimensional technology combinations.3. Supply Chain Resilience: Regionalization and Green Transformation Reshape Global PatternsGeopolitical and sustainability demands are driving the supply chain from “global efficiency first” to “regional resilience first”:The U.S. CHIPS Act and the EU “Chip Act” are promoting domestic capacity building, evolving global wafer manufacturing from “80% in East Asia” to a “tri-polar balance” (each accounting for 20%+ in the U.S., Europe, and Asia).China is accelerating the localization of mature processes above 28nm, with equipment sales revenue expected to grow by 35% in 2024, focusing on long-cycle demand areas such as automotive and industrial control.Data centers account for 8% of global electricity consumption (by 2030), driving investments in liquid cooling technology and low-carbon processes like GaN/SiC; ESG standards directly impact capital expenditure decisions.The U.S. has imposed tariffs of 10%-30% on semiconductor-related products, leading over 60% of manufacturers to initiate supplier substitution plans, increasing supply chain costs and extending delivery cycles.Strategic Insight: Companies need to build a “dual-track supply chain” of “globalization + regionalization”, maintaining global division of labor for high-end chips while achieving regional self-sufficiency for basic chips.4. Corporate Strategy: Dynamic Capabilities and Cycle-Crossing StrategiesIn the face of increasing cyclical volatility, companies need to shift from static planning to dynamic capability building:Adopting a “hybrid IDM + Foundry model” (such as Intel’s foundry services and Samsung’s expansion of diverse customers), retaining core processes while outsourcing non-advantageous segments.Investing in intelligent capacity control systems to adjust production lines based on real-time demand data (e.g., reducing inventory turnover days from 150 to within 100 days through IoT).Leading companies will allocate 30% of R&D investment to disruptive technologies (such as CFET, carbon nanotubes), while continuing to pursue mature paths like FinFET scaling.(e.g., from 2025-2027): Expanding production of high-margin products (HBM/AI accelerators), locking in long-term agreements (LTA) to resist price fluctuations.(e.g., from 2030-2032): Increasing R&D mergers and acquisitions (with leading companies increasing R&D by 15% in 2023), laying out next-generation technologies (such as robotic semiconductors).Collaborating with end manufacturers (e.g., TSMC binding CoWoS with NVIDIA) to quickly capture demand rebounds.Conclusion: Key Windows and Decisive Factors for the Next DecadeThe semiconductor market is expected to exceed $1 trillion by 2030 and reach $1.2 trillion by 2035, but the growth structure will be restructured—automotive electronics will account for 20%, and AI hardware will have a CAGR of 27%.System-level innovation (Chiplet/advanced packaging) capabilities will replace single process leadership.Regional resilience will become a core competitiveness beyond cost.Accurately grasping the AI computing power window from 2025-2027 and the robotics window from 2030-2032 will be crucial.The competition in the next decade will be a comprehensive contest of technological foresight, supply chain resilience, and cyclical insight. Only by internalizing cyclical patterns into strategic rhythms can one continuously capture value dividends in the trillion-dollar market.